BackgroundRelapse is distressingly common after the first episode of psychosis, yet it is poorly understood and difficult to predict. Investigating changes in cognitive function preceding relapse may provide new insights into the underlying mechanism of relapse in psychosis. We hypothesized that relapse in fully remitted first-episode psychosis patients was preceded by working memory deterioration.MethodVisual memory and verbal working memory were monitored prospectively in a 1-year randomized controlled trial of remitted first-episode psychosis patients assigned to medication continuation (quetiapine 400 mg/day) or discontinuation (placebo). Relapse (recurrence of positive symptoms of psychosis), visual (Visual Patterns Test) and verbal (Letter–Number span test) working memory and stressful life events were assessed monthly.ResultsRemitted first-episode patients (n = 102) participated in the study. Relapsers (n = 53) and non-relapsers (n = 49) had similar baseline demographic and clinical profiles. Logistic regression analyses indicated relapse was associated with visual working memory deterioration 2 months before relapse [odds ratio (OR) 3.07, 95% confidence interval (CI) 1.19–7.92, P = 0.02], more stressful life events 1 month before relapse (OR 2.11, 95% CI 1.20–3.72, P = 0.01) and medication discontinuation (OR 5.52, 95% CI 2.08–14.62, P = 0.001).ConclusionsVisual working memory deterioration beginning 2 months before relapse in remitted first-episode psychosis patients (not baseline predictor) may reflect early brain dysfunction that heralds a psychotic relapse. The deterioration was found to be unrelated to a worsening of psychotic symptoms preceding relapse. Testable predictors offer insight into the brain processes underlying relapse in psychosis.